Anthony, I'm fairly sure what you're calling a "parity" bit is what Jon and I are calling a "toggle" bit. In fact I think the Polycom has both a toggle bit and a parity bit (actually a toggle bit and probably two parity bits) but only the toggle bit is significant for the problem in using learned signals.
I did a search and found only one thread that indicates Polycom has a toggle bit
[Link: remotecentral.com]and it only describes the symptom of the toggle bit, no one seems to have explained the symptom in that thread nor mentioned toggle (or parity) bits. However that symptom description does increase my confidence in the guess that this signal does have a toggle bit.
Jon, If those signals at Premise were learned signals then I certainly would expect the toggle bit to vary as you said. I'm leaping to a lot of conclusions here on scant evidence and one of those conclusions is that those signals at Premise were derived from a program based on a protocol description, rather than directly learned from a Polycom remote. I'd also guess that the last bit of each group of eight is even parity for that group rather than the 16'th bit being even parity for all 16 bits. Without seeing a different device code we couldn't really know that and it wouldn't matter for generating the signal.